PG-WM-LATV: a prior-guided weighted multiscale local anisotropic total variation algorithm for missing wedge...

Area of Science:

• Imaging Science
• Computational Science
• Materials Science

Background:

• Computed tomography (CT) is crucial for materials science, biomedical research, and industrial inspection.
• Limited-angle data acquisition, due to hardware or specimen constraints, leads to the "missing wedge" phenomenon, compromising image fidelity.
• Existing methods like local anisotropic total variation (LATV) improve detail preservation but can be further optimized.

Purpose of the Study:

• To develop and evaluate a novel algorithm, prior-guided weighted multiscale LATV (PG-WM-LATV), for accurate CT reconstruction from limited-angle data.
• To enhance the restoration of structural information lost due to the missing wedge phenomenon.
• To improve the overall structural fidelity of reconstructed CT images.

Main Methods:

• Proposed the PG-WM-LATV algorithm incorporating a weighted data fidelity term (Parker function) and a multi-directional regularization term using prior information.
• Implemented a multiscale post-processing strategy with Gaussian pyramid decomposition for hierarchical image analysis.
• Applied alternating frequency-domain compensation, local projection-domain restoration, and directional total variation (TV) regularization at each scale.

Main Results:

• PG-WM-LATV demonstrated superior performance compared to TV, RwATV, and LATV across various limited-angle ranges and Poisson noise levels.
• The algorithm effectively restored information in regions affected by the missing wedge.
• Qualitative and quantitative analyses confirmed significantly enhanced structural fidelity in PG-WM-LATV reconstructed images.

Conclusions:

• The PG-WM-LATV algorithm offers a robust solution for limited-angle CT reconstruction, effectively addressing the missing wedge problem.
• The proposed method achieves adaptive fusion of global structure and local details, leading to high-fidelity reconstructions.
• PG-WM-LATV shows significant potential for applications requiring accurate 3D structural information from incomplete CT datasets.